[OE-core] [PATCH] tzdata_2011n: provide pkg_postinst and timezone defaults

[Andrea Adami]

On Mon, Jan 23, 2012 at 5:49 PM, Mark Hatle <mark.hatle at windriver.com> wrote:
> On 1/23/12 8:21 AM, Andrea Adami wrote:
>>
>> On Sun, Jan 15, 2012 at 12:22 PM, Koen Kooi<koen at dominion.thruhere.net>
>>  wrote:
>>>
>>>
>>> Op 15 jan. 2012, om 12:17 heeft Andrea Adami het volgende geschreven:
>>>
>>>> On Sat, Jan 14, 2012 at 4:10 PM, Koen Kooi<koen at dominion.thruhere.net>
>>>>  wrote:
>>>>>
>>>>>
>>>>> Op 14 jan. 2012, om 14:07 heeft Andrea Adami het volgende geschreven:
>>>>>
>>>>>> * Take the extra bits in do_install from meta-openembedded
>>>>>> * and add pkg_postinst script (adapted from Gentoo).
>>>>>> * Default to 'Universal' timezone, present in the base package
>>>>>> * and allow override using DEFAULT_TIMEZONE variable.
>>>>>> * /etc/localtime and /etc/timezone are purposedly not packaged.
>>>>>> * Bump PR
>>>>>>
>>>>>> Signed-off-by: Andrea Adami<andrea.adami at gmail.com>
>>>>>> ---
>>>>>> meta/recipes-extended/tzdata/tzdata_2011n.bb |   55
>>>>>> +++++++++++++++++++++++++-
>>>>>> 1 files changed, 53 insertions(+), 2 deletions(-)
>>>>>>
>>>>>> diff --git a/meta/recipes-extended/tzdata/tzdata_2011n.bb
>>>>>> b/meta/recipes-extended/tzdata/tzdata_2011n.bb
>>>>>> index de9281a..862a196 100644
>>>>>> --- a/meta/recipes-extended/tzdata/tzdata_2011n.bb
>>>>>> +++ b/meta/recipes-extended/tzdata/tzdata_2011n.bb
>>>>>> @@ -5,7 +5,7 @@ LICENSE = "PD"
>>>>>> LIC_FILES_CHKSUM =
>>>>>> "file://asia;beginline=2;endline=3;md5=06468c0e84ef4d4c97045a4a29b08234"
>>>>>> DEPENDS = "tzcode-native"
>>>>>>
>>>>>> -PR = "r0"
>>>>>> +PR = "r1"
>>>>>>
>>>>>> RCONFLICTS= "timezones timezone-africa timezone-america
>>>>>> timezone-antarctica \
>>>>>>              timezone-arctic timezone-asia timezone-atlantic \
>>>>>> @@ -19,6 +19,8 @@ SRC_URI[tzdata.sha256sum] =
>>>>>> "a343e542486b2b8ebdeca474eed79f1c04f69420ca943c2b9bd
>>>>>>
>>>>>> S = "${WORKDIR}"
>>>>>>
>>>>>> +DEFAULT_TIMEZONE ?= "Universal"
>>>>>> +
>>>>>> TZONES= "africa antarctica asia australasia europe northamerica
>>>>>> southamerica  \
>>>>>>          factory solar87 solar88 solar89 etcetera backward systemv \
>>>>>>         "
>>>>>> @@ -41,7 +43,56 @@ do_install () {
>>>>>>         # libc is removing zoneinfo files from package
>>>>>>         cp -pP "${S}/zone.tab" ${D}${datadir}/zoneinfo
>>>>>>         cp -pP "${S}/iso3166.tab" ${D}${datadir}/zoneinfo
>>>>>> -     chown -R root:root ${D}
>>>>>> +
>>>>>> +        # Install defaults
>>>>>> +        install -d ${D}${sysconfdir}
>>>>>> +        echo ${DEFAULT_TIMEZONE}>  ${D}${sysconfdir}/timezone
>>>>>> +        cp -pP ${S}/${datadir}/zoneinfo/${DEFAULT_TIMEZONE}
>>>>>> ${D}${sysconfdir}/localtime
>>>>>
>>>>>
>>>>> Shouldn't that be an 'ln -sf' instead of a 'cp'?
>>>>>
>>>> No, I don't think so.First reason is /usr may not be mounted, maybe
>>>> living on another device..
>>>> See first Google link about this:
>>>> http://www.redhat.com/archives/rhl-devel-list/2006-April/msg00017.html
>>>> Then, as you see, the postinst takes care of both cases.
>>>
>>>
>>> That only applies to non-UTC times in RTC, which we don't use. And
>>> furthermore, no machine has /usr on a seperate partition since support for
>>> that is broken.
>>
>>
>> I try to bump the thread before the sources get renamed ;)
>>
>> About separation of exec_prefix, pls see
>>
>>
>> http://lists.linuxtogo.org/pipermail/openembedded-core/2012-January/015672.html
>>
>> http://lists.linuxtogo.org/pipermail/openembedded-core/2012-January/015771.html
>>
>> So, what is the position to take?  Symlink or copy?
>
>
> On other systems I've used.. the rule has been: hardlink (same disk) or
> copy. Since the symlink would point into /usr and that should usually be
> avoided (as mentioned above).
>
> I thought there used to be a tool that would do the hardlink, if possible
> and then copy if not.. but I'm not sure if that is even reasonable with our
> image creation... (so copy is likely best answer then)
>
> --Mark
>
>
>> Regards
>>
>> Andrea
>>
>>
>>> _______________________________________________
>>> Openembedded-core mailing list
>>> Openembedded-core at lists.openembedded.org
>>> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
>>
>>
>> _______________________________________________
>> Openembedded-core mailing list
>> Openembedded-core at lists.openembedded.org
>> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
>
>
>
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core at lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

Please take on hold this patch.
Further inspection of do_rootfs logs reveals that the postinst is run
during that task and the code tries to write /etc/localtime on host,
fails and exits 1.

I'll add a check like

pkg_postinst_${PN}() {
if test "x$D" != "x"; then
	exit 1
else

and send a revised patch in a few hours.

Regards

Andrea

P.S.
This is a low hanging fruit wrt recipes duplication between oe-core and meta-oe.
Please add your review/comments, thanks.